3 ms·

wow - do you mind sharing any links to a specific setup? Also whats the biggest model anybody has run on this?

• lossolo 1y ago
  Unfortunately, you will not be able to run any model on this that is comparable to the Claude models.
• paxys 1y ago
  You can run a decent model on it, say highly quantized Qwen or Deepseek R1 getting 5-10 tokens/sec output, but it will be nothing in comparison to a commercial offering like Claude, o3 or Gemini. For that you need a datacenter-class GPU going for $50K-100K a pop.
  • mtkd 1y ago
    But a small collective running that box, especially spanning timezones, could potentially be a viable alternative or will be soon -- with obv privacy gains too
• icelancer 1y ago
  Every model you run on that setup will be at best half as good as Sonnet 4.